Chinese and Vietnamese neural machine translation method fusing zero pronouns and chapter information

Abstract

The invention relates to a Chinese and Vietnamese neural machine translation method fusing zero pronouns and chapter information, and belongs to the technical field of natural language processing. The method comprises the following steps: constructing a middle-to-cross-English three-language aligned chapter data set, and performing zero pronoun classification marking on middle-to-cross data; respectively acquiring bilingual features of a source statement and a context by using a self-attention mechanism; the source statement and the context features are pooled and linked, and syntactic component classification of null pronouns is carried out; and the target statement is predicted through the two attention sub-layers by using the source statement and the context features. A joint learning mode is adopted, and parameters of the main task model and the auxiliary model are learned and updated at the same time. And combining the classification task and the translation task. And chapter information is added in the classification task, so that the zero pronoun classification accuracy is improved. And meanwhile, the chapter information can also effectively improve the translation task performance. By fusing the null pronouns and the chapter information, the Chinese and Vietnamese neural machine translation performance is effectively improved.

Description

technical field [0001] The invention relates to a Chinese-Vietnamese neural machine translation method integrating zero pronouns and textual information, and belongs to the technical field of natural language processing. Background technique [0002] The exchanges between China and Vietnam are getting closer and closer, and the demand for Chinese-Vietnamese translation technology is growing. The research on translation technology in low-resource scenarios such as Chinese-Vietnamese is getting better and better. However, the translation technology researched at this stage is more aimed at the usage scenarios of formal style, such as the translation of news texts and official documents. For informal styles, such as online comments, spoken daily conversations and other usage scenarios, the translation performance is obviously insufficient under the same translation model. The main reason for the poor translation performance is that in spoken language and daily dialogue scenar...

AI Technical Summary

Problems solved by technology

Although Transformer can use the multi-head attention mechanism to capture more semantic information, it often only translates simple parts of the omitted pronouns

Images